一种局部视角的类别近似质量属性约简加速方法  

A LOCAL VIEW BASED ACCELERATION APPROACH FOR CLASS-SPECIFIC APPROXIMATE QUALITY BASED ATTRIBUTE REDUCTION

作　　者：李智远 饶先胜 宋晶晶[2] 杨习贝[2] Li Zhiyuan;Rao Xiansheng;Song Jingjing;Yang Xibei(Kewen College,Jiangsu Normal University,Xuzhou 221116,Jiangsu,China;School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212003,Jiangsu,China)

机构地区：[1]江苏师范大学科文学院,江苏徐州221116 [2]江苏科技大学计算机学院,江苏镇江212003

出　　处：《计算机应用与软件》2021年第11期249-254,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61572242,61906078)。

摘　　要：在利用贪心搜索算法求解类别近似质量约简的过程中,类别近似质量的计算是评估属性重要度的重要步骤,需要考虑论域中所有样本的邻域与当前决策类之间的包含关系。为降低求解类别近似质量约简的时间消耗,从局部的视角出发,提出一种用于求解类别近似约简的加速方法。该方法在计算类别近似质量时仅考虑当前决策类中的样本而不是论域中所有的样本,通过减少计算规模以加快约简求解的过程。在8个UCI数据集上的实验结果表明,该方法在不改变约简结果的情况下,能显著降低求解约简的时间消耗。In the process of finding class-specific approximate quality reduction with greedy searching algorithm, the deriving of class-specific approximate quality is the key step for evaluating the significances of candidate attributes, and the neighborhoods of all samples in universe should be considered to induce class-specific approximate quality. In order to reduce the time consumption for finding class-specific approximate quality reduction, from the local perspective, an acceleration approach for solving this reduction is proposed. This approach was designed by only considering the samples in the current decision class instead of the whole universe when computing class-specific approximate quality, and it reduced the calculation scale to speed up the process of searching reduction. The experimental results over 8 UCI data sets show that, the proposed acceleration approach can significantly reduce the elapsed time of searching reduction without changing the reduction results.

关 键 词：加速方法 属性约简 类别近似质量 局部视角 邻域粗糙集 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]